David Josephs

About Me

I am a:

• Aspiring data scientist

• Open source contributor (author of 1.5 R packages and 1 python package, as well as 2 somewhat widely used covid research packages)

• Chess player

• Vim and Linux enthusiast

• TA to a graduate level data science course

My blog

Data Science Blog:

Recent Articles:

Important note, I am having some serious issues with systme fonts lately, therefore some graphics may look a bit funny…

Interpretable Machine Learning: Part 1

Building a Reproducible Data Science Environment with Nix

Forecasting Air Pollution Part 1: Preprocessing

Forecasting Air Pollution Part 2: Univariate Exploratory Data Analysis

Chess Blog

Work in progress

Things I love

• Python
• R
• Nix
• Long walks
• Chess
• Writing
• Explainable Models
• Parametrized Deep Learning
• Time Series
• My dog

Things I am doing

• Dropping everything to work on COVID-19
  • Direct contributions:
    • Tools for parsing COVID papers
    • Tools for covid modeling
• Working on my master’s thesis (Fast Classification of sEMG signal):
  • Fast classification of electrical signals in muscles. The idea is we can quickly classify a broad range of motions in very short time windows, using deep techniques on the near raw electrical signals. The usage of this would be mainly in prosthetics which predict gestures based on what the user wants, and body controlled electrical arms. Our classifier is able to classify 12 fine finger movements, 17 wrist movements, 23 functional motions, and rest with relatively high accuracy in less than 300 milliseconds, which makes it inherently useful. However, in the future I believe that an unsupervised or reinforcement based approach would be more useful, the presence of classes at all is a hindrance on the functionality of these models.
• Helping graduate students succeed and understand the entire data science workflow from data attainment to deployment (head TA for a survey of data science course, and weekend lecturer for an advanced machine learning course)
• Maintaining and growing my first large python library
• Looking for projects to keep me busy. Ideas:
  • More blog posts!
  • Interpretable Machine Learning
  • Build a VAE picture based “restaurant”

Things I have done

• Worked for NASA
  • Acted as translator, astronomer, and lead scout for the ultima thule mission
  • Conducted televised interviews in Argentina
• Performed extensive time series analysis of \(PM_{2.5}\) content in Beijing, using models ranging from ARIMA and VAR to LSTMs, and ensembling methods
• Conducted ARM using R on census data, yielding unsurprising results
• Extensively done classification on census data using python
• Built a snowflake schema data store at treehouse
  • Developed a SKU system for all products, ranging from paint to refrigerators
• Managed data analytics for a local election at Uptown Dallas, Inc